estReg: Bayesian estimation in mixed nonlinear regression models

In SimoneHermann/hierRegSDE:

Description

Bayesian estimation of the random effects φ_j in the mixed nonlinear regression model y_{ij}= f(φ_j, t_{ij}) + ε_{ij}, ε_{ij}~N(0,γ^2*s^2(t_{ij}), φ_j~N(μ, Ω) and the parameters μ, Ω, γ^2.

Usage

estReg(t, y, prior, start, fODE, sVar, ipred = 1, cut, len = 1000,
  Omega = "diag", mod = c("Gompertz", "logistic", "Weibull", "Richards",
  "Paris", "Paris2"), propPar = 0.02)

Arguments

t	vector of observation times
y	matrix of the M trajectories
prior	list of prior parameters - list(m, v, priorOmega, alpha, beta)
start	list of starting values
fODE	regression function
sVar	variance function
ipred	which of the M trajectories is the one to be predicted
cut	the index how many of the ipred-th series are used for estimation
len	number of iterations of the MCMC algorithm
Omega	structure of the variance matrix Omega of the random effects, diagonal matrix, otherwise inverse wishart distributed
mod	model out of Gompertz, Richards, logistic, Weibull, only used instead of fODE
propPar	proposal standard deviation of phi is |start$mu|*propPar

Value

phi	samples from posterior of φ
mu	samples from posterior of μ
Omega	samples from posterior of Ω
gamma2	samples from posterior of γ^2

Examples

mod <- "Gompertz"
fODE <- getFun("ODE", mod)
mu <- getPar("ODE", mod, "truePar")
n <- 5
parameters <- defaultPar(mu, n)
y <- drawData("ODE", fODE, parameters)
t <- parameters$t
prior <- getPrior(mu, parameters$gamma2)
start <- getStart(mu, n)
chains <- estReg(t, y, prior=prior, start=start, fODE=fODE)
plot(phi_ij(chains$phi, 1, 1), type="l")
plot(chains$gamma, type="l"); abline(h=parameters$gamma2, col=2)

SimoneHermann/hierRegSDE documentation built on May 9, 2019, 1:46 p.m.